Back to EveryPatent.com
| United States Patent |
5,029,965
|
|
Tan
|
July 9, 1991
|
Method for aligning an optical fiber to an active device
Abstract
A method for locating an optimum alignment of an optical fiber to an active
device uses response surface approximations. In an X-Y coordinate plane
perpendicular to the optical fiber, measurements at a first series of
alignments are made from which the path of steepest ascent of light
transfer between the optical fiber and the active device is approximated
by first order design model. Further measurements are made within a more
limited region determined by the first order design model approximation.
From the further measurements second order modelling is done to
approximate the alignment resulting in maximum light transfer between the
optical fiber and the active device.
| Inventors:
|
Tan; Yoke T. (Toa Payoh, Lorong 1, #03-584, Singapore 1231 Blk 165, SG)
|
| Appl. No.:
|
440000 |
| Filed:
|
November 21, 1989 |
| Current U.S. Class: |
385/90 |
| Intern'l Class: |
G02B 006/42 |
| Field of Search: |
364/194
350/96.15,96.20,320
357/17,19,30
250/227.29
|
References Cited
U.S. Patent Documents
| 4787698 | Nov., 1988 | Lyons et al. | 350/96.
|
| 4802726 | Feb., 1989 | Palmquist et al. | 350/96.
|
| 4854667 | Aug., 1989 | Ebata et al. | 350/96.
|
| 4887882 | Dec., 1989 | Mousseaux et al. | 350/96.
|
Other References
Box, Hunter & Hunter, Statistics for Experimenters, Chapters 14 & 15.
|
Primary Examiner: Gonzalez; Frank
Claims
I claim:
1. A method for locating an optimum alignment of an optical fiber to an
active device in an X-Y coordinate plane perpendicular to the optical
fiber, the optimum alignment being at X-Y coordinates where an optimal
transfer of light between the active device and the optical fiber takes
place, the method comprising the steps of:
(a) aligning the optical fiber to the active device at a first series of
X-Y coordinates and taking first measurements which indicate the relative
amount of light transferred between the optical fiber and the active
device at each of the first series of X-Y coordinates;
(b) making a first order design model approximation of the path of steepest
ascent of the amount of transferred light relative to selected X-Y
coordinates, the first order design model approximation being based on the
first measurements;
(c) taking second measurements of the amount of transferred light when the
optical fiber is aligned with the active device at a second series of X-Y
coordinates along the approximation of the path of steepest ascent;
(d) taking third measurements of the amount of transferred light when the
optical fiber is aligned with the active device at a third series of X-Y
coordinates, the third series of X-Y coordinates being in a region which
includes the X-Y coordinate in the second series of X-Y coordinates at
which the maximum amount of transferred light was measured;
(e) making a second order composite design model approximation of the
approximate optimal alignment X-Y coordinate at which the maximum amount
of light is transferred, the second order composite design model
approximation being based on the third measurements;
(f) taking fourth measurements of the amount of transferred light when the
optical fiber is aligned with the active device at a fourth series of X-Y
coordinates near the approximate optimal alignment X-Y coordinate; and,
(g) selecting as the optimum alignment of the optical fiber to the active
device an X-Y coordinate from the fourth series of X-Y coordinates at
which the maximum amount of transferred light was measured.
2. A method as in claim 1 wherein the first order design model
approximation of the steepest past ascent is calculated using the
following equation:
Po=b0=b1X1=b2Y1;
where Po is proportional to the amount of transferred light, X1 is
proportional to the X coordinate, Y1 is proportional to the Y coordinate
and b0, b1 and b2 are coefficients.
3. A method as in claim 1 wherein the second order composite design model
approximation of the steepest past ascent is calculated using the
following equation:
Po=b0=b1X1=b2Y1=b11(X1.sup.2)=b22(Y1.sup.2)=b12(X1Y1)
where Po is proportional to the amount of transferred light, X1 is
proportional to the X coordinate, Y1 is proportional to the Y coordinate
and b0, b1, b2, b11, b22 and b12 are coefficients.
4. A method for locating an optimum alignment of an optical fiber to an
active device, the optimum alignment being at X-Y-Z coordinates where an
optimal transfer of light between the active device and the optical fiber
takes place, the method comprising the steps of:
(a) aligning the optical fiber to the active device at a first series of
X-Y coordinates in an X-Y coordinate plane perpendicular to the optical
fiber and taking first measurements which indicate the relative amount of
light transferred between the optical fiber and the active device at each
of the first series of X-Y coordinates;
(b) making a first order design model approximation of the path of steepest
ascent of the amount of transferred light relative to selected X-Y
coordinates, the first order design model approximation being based on the
first measurements;
(c) taking second measurements of the amount of transferred light when the
optical fiber is aligned with the active device at a second series of X-Y
coordinates along the approximation of the path of steepest ascent;
(d) taking third measurements of the amount of transferred light when the
optical fiber is aligned with the active device at a third series of X-Y
coordinates, the third series of X-Y coordinates being in a region which
includes the X-Y coordinate in the second series of X-Y coordinates at
which the maximum amount of transferred light was measured;
(e) making a second order design model approximation of the approximate
peak alignment X-Y coordinate at which the maximum amount of light is
transferred, the second order design model approximation being based on
the third measurements;
(f) taking fourth measurements of the amount of transferred light when the
optical fiber is aligned with the active device at a fourth series of X-Y
coordinates near the approximate peak alignment X-Y coordinate;
(g) selecting as a peak X-Y coordinate alignment of the optical fiber to
the active device an X-Y coordinate from the fourth series of X-Y
coordinates at which the maximum amount of transferred light was measured;
(h) taking fifth measurements of the amount of transferred light when the
optical filter is aligned at a fifth series of coordinate points in which
the Z coordinate is varied and X-Y coordinates are at the peak X-Y
coordinate alignment;
(i) selecting as the optimum Z coordinate the Z coordinate in the fifth
series of coordinate points at which the maximum amount of transferred
light was measured;
(j) taking sixth measurements of the amount of transferred light when the
optical fiber is aligned with the active device at a sixth series of
coordinates, the sixth series of coordinates being in plane of the optimum
Z coordinate and in a region which includes the X-Y peak coordinate;
(k) making a second order design model approximation of the approximate
optimal alignment X-Y coordinate at which the maximum amount of light is
transferred, the second order design model approximation being based on
the sixth measurements;
(l) taking seventh measurements of the amount of transferred light when the
optical fiber is aligned with the active device at a seventh series of
coordinates near the approximate optimal alignment X-Y coordinate; and,
(g) selecting as a optimal coordinate alignment of the optical fiber to the
active device an X-Y-Z coordinate from the seventh series of coordinates
at which the maximum amount of transferred light was measured.
5. A method as in claim 4 wherein the first order design model
approximation of the steepest past ascent is calculated using the
following equation:
Po=b0=b1X1=b2Y1
where Po is proportional to the amount of transferred light, X1 is
proportional to the X coordinate, Y1 is proportional to the Y coordinate
and b0, b1 and b2 are coefficients.
6. A method as in claim 4 wherein the second order design model
approximation of the steepest past ascent is calculated using the
following equation:
Po=b0=b1(X1)=b2(Y1)=b11(X1.sup.2) =b22(Y1.sup.2)=b12(X1Y1);
where Po is proportional to the amount of transferred light, X1 is
proportional to the X coordinate, Y1 is proportional to the Y coordinate
and b0, b1, b2, b11, b22 and b12 are coefficients.
7. A method for locating an optimum alignment of an optical fiber to an
active device, the optimum alignment being at X-Y-Z coordinates where an
optimal transfer of light between the active device and the optical fiber
takes place, the method comprising the steps of:
(a) aligning the optical fiber to the active device at a first series of
X-Y-Z coordinates and taking first measurements which indicate the
relative amount of light transferred between the optical fiber and the
active device at each of the first series of X-Y-Z coordinates;
(b) making a first order design model approximation of the path of steepest
ascent of the amount of transferred light relative to selected X-Y-Z
coordinates, the first order design model approximation being based on the
first measurements;
(c) taking second measurements of the amount of transferred light when the
optical fiber is aligned with the active device at a second series of
X-Y-Z coordinates along the approximation of the path of steepest ascent;
(d) taking third measurements of the amount of transferred light when the
optical fiber is aligned with the active device at a third series of X-Y-Z
coordinates, the third series of X-Y-Z coordinates being in a region which
includes the X-Y-Z coordinate in the second series of X-Y-Z coordinates at
which the maximum amount of transferred light was measured;
(e) making a second order composite design model approximation of the
approximate optimal alignment X-Y-Z coordinate at which the maximum amount
of light is transferred, the second order composite design model
approximation being based on the third measurements;
(f) taking fourth measurements of the amount of transferred light when the
optical fiber is aligned with the active device at a fourth series of
X-Y-Z coordinates near the approximate optimal alignment X-Y-Z coordinate;
and,
(g) selecting as the optimum alignment of the optical fiber to the active
device an X-Y-Z coordinate from the fourth series of X-Y-Z coordinates at
which the maximum amount of transferred light was measured.
8. A method as in claim 7 wherein the first order design model
approximation of the steepest past ascent is calculated using the
following equation:
Po=b0=b1X1=b2Y1=b3Z1;
where po is proportional to the amount of transferred light, X1 is
proportional to the X coordinate, Y1 is proportional to the Y coordinate,
Z1 is proportional to the Z coordinate and b0, b1, b2 and b3 are
coefficients.
9. A method as in claim 7 wherein the second order composite design model
approximation of the steepest past ascent is calculated using the
following equation:
Po=b0=b1X1=b2Y1=b3Z3=b12(X1Y1)=b13(X1Z1) =b23(Y1Z1)=b123X1Y1Z1;
where Po is proportional to the amount of transferred light, X1 is
proportional to the X coordinate, Y1 is proportional, to the Y coordinate,
Z1 is proportional to the Z coordinate and b0, b1, b2, b3, b12, b13, b23
and b123 are coefficients.
Description
BACKGROUND
The present invention concerns the alignment of an optical fiber to a light
source or to a light detector before permanent attachment.
When attaching an optical fiber to a light source, for example a laser
diode or a light emitting diode (LED), or to a light detector, for example
a PIN detector, alignment is critical. Alignment is a complicated process
due to the fact that the narrow diameter of an optical fiber and the
limited size of a chip which serves as a light source or light detector.
For example, the diameter of an optical fiber is typically 62.5
micrometers. A PIN detector typically has an active area with a diameter
of approximately 60 micrometers. An LED typically has a light emission
area with a diameter of 20 micrometers. Therefore, accuracy of alignment
within a micrometer is essential. A misalignment by even a couple of
micrometers can cause a variation in optical power budget loss of a few
dBm.
Until recently, alignment was generally performed manually. Highly
experienced operators would use manual positioning stages, optical power
meters and large doses of patience to align an optical fiber for
attachment to a light source or a light detector. More recently automated
systems have been used for such alignment. A typical automated alignment
system usually includes mechanical, electrical and electro-optical
components and instrumentations controlled by a computer system.
In a typical automated alignment system either a light source (e.g. LED) or
a light detector (e.g. PIN detector) is assembled on a circular header.
The resultant assembly is then tested. After testing, the assembly is
aligned to the fiber optical connector receptacle. Once aligned, the fiber
optical connector receptacle is permanently attached and sealed to the
assembly in the aligned position to form an optical subassembly (OSA).
Typically, alignment of the assembly of the fiber optical receptacle is
done as follows. A fiber ferrule is plugged into the connector receptacle
to monitor resultant light power from a light source or to measure
resultant current from a light detector. A header portion of the optical
fiber is then moved in an X-Y plane parallel to the plane in which the
emission area or the detection area resides. The header portion is left at
the X-Y coordinates which result in the maximum output power. The port cap
or connector receptacle is then moved along a Z-axis perpendicular to the
plane in which the emission area or detection area resides. This is done
to focus the light beam in order to find the location along the Z-axis
where maximum output power is reached. Then the header is again varied in
the X-Y plane to find the position of maximum power. At this point, the
OSA is ready for permanent attachment.
SUMMARY OF THE INVENTION
In accordance with the preferred embodiment of the present invention a
method for locating an optimum alignment of an optical fiber to an active
device is presented. The optical fiber is aligned to the active device at
a first series of X-Y coordinates. At each of the series of X-Y
coordinates first measurements are taken which indicate the relative
amount of light transferred between the optical fiber and the active
device.
From the first measurements a first order design model is made which
approximates the path of steepest ascent of the amount of transferred
light relative to selected X-Y coordinates. Second measurements are taken
for a second series of X-Y coordinates which are along the approximated
path of steepest ascent.
Third measurements are taken of the amount of transferred light when the
optical fiber is aligned with the active device at a third series of X-Y
coordinates. The third series of X-Y coordinates are in a region which
includes the X-Y coordinate in the second series of X-Y coordinates at
which the maximum amount of transferred light was measured.
A second order composite design model approximation is made of the
approximate optimal alignment X-Y coordinate at which the maximum amount
of light is transferred. The second order composite design model
approximation is based on the third measurements.
Fourth measurements are taken of the amount of transferred light when the
optical fiber is aligned with the active device at a fourth series of X-Y
coordinates near the approximate optimal alignment X-Y coordinate. The
optimum alignment of the optical fiber to the active device an X-Y
coordinate is selected from the fourth series of X-Y coordinates at which
the maximum amount of transferred light was measured.
The method of the present invention allows for alignment of an optical
fiber to an active device with significantly fewer measurements than, for
example, the prior art one-variable-at-a-time method. Since significantly
fewer measurements are made, the alignment can be performed significantly
faster.
The same basic method may be used to align the optical fiber to the active
device using three coordinate axes instead of two.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 shows a block diagram of an automated aligner according to the
preferred embodiment of the present invention.
FIG. 2 shows an optical subassembly (OSA) in accordance to the preferred
embodiment of the present invention.
FIG. 3 shows the OSA in FIG. 2 used in conjunction with a pin diode.
FIG. 4 shows the OSA in FIG. 2 used in conjunction with a light emitting
diode (LED).
FIG. 5 through FIG. 8 show screen displays resulting from a prior art
method for aligning an optical subassembly.
In FIG. 9, alignment points at which response is measured in a first order
design model approximation are shown on a graph representing the
responsiveness of a pin diode in detecting light signals in accordance
with the preferred embodiment of the present invention.
FIG. 10 is a graph which shows estimated response and measured response of
a pin diode variously aligned in the X-Y plane in accordance with the
preferred embodiment of the present invention.
In FIG. 11, alignment points at which response is measured in a second
order composite design model approximation are shown on the graph shown in
FIG. 9 in accordance with the preferred embodiment of the present
invention.
DESCRIPTION OF THE PREFERRED EMBODIMENT
In FIG. 1 a block diagram of automate aligner is shown. An alignment and
sealing station 3 is used to perform actual mechanical alignment of an
optical fiber connector to a chip. The alignment is changed by means of
motors which are able to independently change the alignment along the
X-axis, Y-axis and Z-axis. A motor 15 changes the alignment along the
X-axis, a motor 16 changes the alignment along the Y-axis, and a motor 17
changes the alignment along the Z-axis. Alignment and sealing station 3
also, upon completion of alignment, performs the permanent affixing and
sealing of the optical fiber connector to the chip. The result is a
finished optical subassembly (OSA).
An instrumentation module 2, through a data path 6, provides power and
control signals to a unit 18 which is aligned within alignment and sealing
station 3. Instrumentation module 2 also receives from unit 18 through
data path 6 electrical and/or optical feedback.
A computer 1 communicates with instrumentation module 2, through a data
path 5. In the preferred embodiment, computer 1 is a HP model 9000 series
300 computer and data path 5 is implemented using HPIB (or IEEE 488) bus
protocol. Computer 1 makes calculations based on feedback from
instrumentation module 2 in order to determine X, Y, and Z coordinate
positions to be set by alignment and sealing station 3. Computer 1
communicates with a controller 19 within alignment and sealing station 3
by way of a data path 4. In the preferred embodiment, data path 5 is
implemented using HPIB bus protocol, and controller 19 is a controller
model 4000 available from Kensington Laboratories Inc., a U.S. corporation
having a place of business at 750 National Court, Richmond, Calif. 94804.
Based on signals from computer 1, controller 19 sends control signals to
motor 15, motor 16 and motor 17 to direct change of alignment along the
X-axis, Y-axis and Z-axis.
An optical subassembly is shown in FIG. 2. An optical fiber 10 in a housing
of ceramic ferrule is held to a chip 11 by means of a port cap connector
13. A lens 9 focuses light to or from optical fiber 10. A chip, i.e., a
light emitting diode (LED) or a PIN diode is placed on a circular header
-1. A window 8 separates the chip from lens 9. A coordinate axis 14 shows
the direction of an X-coordinate, a Y-coordinate and a Z-coordinate. The
X-Y plane lies perpendicular to optical fiber -0. Alignment is performed
by moving circular header 11 in the X- direction and the Y direction and
by moving port cap connector 13 in the Z -direction. The movement is done
using the motor 15, 16 and 17.
FIG. 3 shows a PIN detector 101 placed on circular header 11. Light from a
light source 102 is transmitted through optical fiber 10, through lens 9
through window 8 to PIN detector 101. PIN detector 101 will generate
current based on the amount of light which reaches the active area. As the
alignment is adjusted PIN detector 101 sends feedback to instruments 2.
This feedback is used by computer 1 to aid in alignment as discussed
further below.
FIG. 4 shows an LED 201 placed on circular header 11. Light from LED 201 is
transmitted through window 8, through lens 9, through optical fiber 10 to
a detector 202. Detector 202 will generate current based on the amount of
light which reaches the active area. As the alignment of the position of
LED relative to lens 9 is adjusted, detector 202 sends feedback to
instruments 2. This feedback is used by computer 1 to aid in alignment.
The method of alignment according to the present invention uses a response
surface methodology as an alignment algorithm to align optical fiber 10 in
the X-Y coordinate plane over a known contour response. The method
according to the present invention is contrasted to a
one-variable-at-a-time strategy for determining the optimum alignment of
the optical fiber in the X-Y coordinate plane.
In the one-variable-at-a-time strategy, as applied to a light detector such
as a PIN detector, light is propagated through optical fiber 10 through
window 8 to active area 12. The current generated by light impacting
active area 12 (referred to as the "response") is monitored. The alignment
of chip 11 to optical fiber 10 is arbitrarily varied along one coordinate
axis, while the alignment of chip 11 to optical fiber 10 is held constant
along the other coordinate axes. A step-by-step methodology for a
one-variable-at-a -time strategy is set out below:
Step 1: The initial X, Y and Z coordinates at the beginning of the
alignment are defined as being at the coordinate position of (0,0,0). A
search window of 200 micrometers in diameter is defined in the X-Y
coordinate plane with its center at the coordinate position of (0,0,0).
The optimum alignment in the X-Y coordinate plane is assumed to be within
this window.
Step 2: Without varying alignment along the Y-coordinate axis or the
Z-coordinate axis the alignment along the X axis is varied at increments
of 5 micrometers starting from the coordinate point (-100,0,0) and
continuing to the coordinate point (100,0,0), where the coordinate points
are given in micrometers. The response at each increment is measured by
evaluating the power output generated by the PIN detector. A curve 21,
shown in FIG. 3, of the response is plotted onto a screen 20 of computer 1
and stored in a data array within a memory of computer 1.
Step 3: From the data array, computer 1 calculates a point 22 where
response is 90% of the maximum response from the left side of curve 21 and
a point 23 where response is 90% of the maximum response from the right
side of curve 21. A midpoint 24 between the 90% response points is chosen
as the peak X-coordinate point. Computer 1 will instruct alignment and
sealing station 3 to go to this coordinate point. The coordinate point may
be labelled (X-peak,0,0).
Step 4: Without varying alignment along the X-coordinate axis or the
Z-coordinate axis, the alignment along the Y axis is varied at increments
of 5 micrometers starting from the coordinate point (X-peak,-100,0) and
continuing to the coordinate point (X-Peak,100,0). The response at each
increment is measured by evaluation the power output generated by the PIN
detector. A curve 31, shown in FIG. 4, of the response is plotted onto
screen 20 of computer 1 and stored in a data array within the memory of
computer 1.
Step 5: From the data array, computer 1 calculates a point 32 where
response is 90% of the maximum response from the left side of curve 31 and
a point 33 where response is 90% of the maximum response from the right
side of curve 31. A midpoint 34 between the 90% response points is chosen
as the peak Y-coordinate point. Computer 1 will instruct alignment and
sealing station 3 to go to this coordinate point. The coordinate point may
be labelled (X-peak,Y-peak,0).
Step 6: Vary the alignment along the Z axis until the light beam from
optical fiber 10 is at peak response power. The coordinate point for this
position is (X-peak,Y-peak,Z-optimum). For step 7 through step 10 below
coordinate point (X-peak,Y-peak,Z-optimum) is re-labelled (0,0,Z-optimum).
Step 7: Without varying alignment along the Y-coordinate axis or the
Z-coordinate axis, the alignment along the X axis is varied at increments
of 5 micrometers starting from the coordinate point (-100,0,Z-optimum) and
continuing to the coordinate point (100,0,Z-optimum). A curve 41, shown in
FIG. 5, of the response is plotted onto screen 20 of computer 1 and stored
in a data array within the memory of computer 1.
Step 8: From the data array, computer 1 calculates a point 42 where
response is 90% of the maximum response from the left side of curve 41 and
a point 43 where response is 90% of the maximum response from the right
side of curve 41. A midpoint 44 between the 90% response points is chosen
as the optimum X-coordinate point. Computer 1 will instruct alignment and
sealing station 3 to go to this coordinate point. The coordinate point may
be labelled (X-optimum,0,Z-optimum).
Step 9: Without varying alignment along the X-coordinate axis or the
Z-coordinate axis the alignment along the Y axis is then varied at
increments of 5 micrometers starting from the coordinate point (X
-optimum,-100,Z-optimum) and continuing to the coordinate point
(X-optimum,100,Z-optimum). The response at each increment is measured by
evaluation the power output generated by the PIN detector. A curve 51,
shown in FIG. 6, of the response is plotted onto screen 20 of computer 1
and stored in a data array within the memory of computer 1.
Step 10: From the data array, computer 1 calculates a point 52 where
response is 90% of the maximum response from the left side of curve 51 and
a point 53 where response is 90% of the maximum response from the right
side of curve 51. A midpoint 54 between the 90% response points is chosen
as the optimum Y-coordinate point. Computer 1 will instruct alignment and
sealing station 3 to go to this coordinate point
(X-optimum,Y-optimum,Z-optimum).
The present invention utilizes a response surface methodology to align
optical fiber 10 to chip 11 in the X-Y plane. In the preferred embodiment
of the present invention this is done using a two squared factorial as an
alignment algorithm for calculating optimum position along the
X-coordinate and the Y-coordinate axis. The response surface methodology
is not used to align optical fiber to chip 11 along the Z-coordinate axis.
For the determination of peak response in the X-Y plane, there are
essentially two basic steps. (1) Measurements are made from which the path
of steepest ascent is determined by first order design model
approximation. (2) Further measurements are made within a more limited
region determined by the first order design model approximation. From the
further measurements second order modelling is done to approximate the
alignment resulting in a maximum response.
The method according to the present invention may be done varying alignment
along three coordinate axes using a two cubed factorial or varying
alignment along two coordinate axes using a two squared factorial.
In the preferred embodiment, a two squared factorial as an alignment
algorithm for calculating optimum position along the X-coordinate and the
Y-coordinate axis was chosen rather than a two cubed factorial along all
three coordinate axes. This is because in the overall system in which the
method was applied, the use of a two cubed factorial requiring alignment
to be varied along the full Z-coordinate axis might result in lens 9 might
be pushed through window 8. Trying to avoid this would result in
arbitrarily limiting range of movement along the Z-coordinate axis, which
could limit data from which the first order and second order
approximations are made. This could result in the optimal response output
not being found. Varying the focal point of lens 9 could solve this
trouble. Applicant instead elected to avoid this problem by using the
response surface methodology to find optimum position along the
X-coordinate axis and the Y-coordinate only, not along the Z-coordinate
axis.
A step-by-step procedure for performing the response surface methodology is
set out below:
Step 1a: The initial X, Y and Z coordinates at the beginning of the
alignment are defined as being at the coordinate position of (0,0,0). The
response will be measured at four coordinate points: (-20,-20,0);
(+20,-20,0); (-20,+20,0); (+20,+20,0) (Values for the coordinate points
are give in micrometers). The four coordinate points are shown in FIG. 9
as points 41, 42, 43 and 44. FIG. 9 is a graph whose elevation lines
indicate the response current generated by pin diode 101 as a function of
the X-Y coordinate position of circular header 11. In order for ease of
calculation, the coordinate points each have a coded form. Respectively
the coded form of the coordinate points are (-1,-1,0); (+1,-1,0); (-1,
+1,0); and (+1,+1,0). A matrix "A" derived from the coordinate points is
as follows:
Matrix A
______________________________________
Matrix A
______________________________________
1.0 -1.0 -1.0
1.0 1.0 -1.0
1.0 -1.0 1.0
1.0 1.0 1.0
______________________________________
A' is defined as the transpose of Matrix A and is as follows:
Matrix A'
______________________________________
Matrix A'
______________________________________
1.0 1.0 1.0 1.0
-1.0 1.0 -1.0 1.0
-1.0 -1.0 1.0 1.0
______________________________________
Step 1b: The response at the four coordinate points is measured and stored
in a matrix P. Table 1 below gives example values for a measured response:
TABLE 1
______________________________________
X-axis X1 Y-Axis Y1 Response (P)
______________________________________
-20 -1 -20 -1 +4.611
+20 +1 -20 -1 +5.390
-20 -1 +20 +1 +3.770
+20 +1 +20 +1 +4.825
______________________________________
Matrix P is as follows:
Matrix P
+4.611
+5.390
+3.770
+4.825
Computer 1 uses the measured response to compute a response surface of a
first order polynomial given to Equation 1 below:
Po=b0=b1X1=b2Y1 (1)
Po is the predicted response at a particular alignment along the X
coordinate axis and the Y coordinate axis. In order to determine
coefficients b0, b1 and b2, computer 1 employs at least squares method
using matrix equation 2 below:
b=[(A'A).sup.-1 ]A'P (2)
For the example of measured responses given in Table 1, values of
coefficients b0, b1 and b2 are calculated to be as follows. b0 equals
+4.611. b1 equals +0.496. b2 equals -0.314.
Step 1c: The above calculations based on the response at the four measured
coordinate points 41, 42, 43, 44, give the a path of steepest ascent of
response 55 shown in FIG. 9. Path 55 is determined by equation 3 below:
Y1=(b2/b1)X1 (3)
In order to confirm estimated values of Po (Po-est) along the path of
steepest ascent as calculated using Equation 1, actual values of Po
(Po-act) are measured. These actual values are measured starting at
coordinate position (0,0,0). The resultant Table 2 below show values of
Po-est calculated from equation 1 as well as typical values of P-act
measured during an actual alignment.
TABLE 2
______________________________________
X-axis
X1 Y-axis Y1 Po-est
Po-act
______________________________________
0 0 0 0 4.61 5.17
+5 +0.25 -3.16 -0.158 4.78 5.38
+10 +0.50 -6.32 -0.316 4.96 5.50
+15 +0.75 -9.48 -0.474 5.13 5.57
+20 +1.00 -12.64 -0.632 5.31 5.55
+25 +1.25 -15.81 -0.790 5.48 5.47
+30 +1.50 -18.97 -0.948 5.65 5.34
+35 +1.75 -22.13 -1.106 5.83 5.13
+40 +2.00 -25.29 -1.264 6.00 4.87
+45 +2.25 -28.45 -1.423 6.17 4.58
+50 +2.50 -31.61 -1.581 6.35 4.22
+55 +2.75 -34.77 -1.739 6.52 3.86
+60 +3.00 -37.93 -1.897 6.70 3.48
+65 +3.25 -41.10 -2.055 6.87 3.10
+70 +3.50 -44.26 -2.213 7.04 2.88
+75 +3.75 -47.42 -2.371 7.22 2.70
______________________________________
A graph of predicted and actual response is shown in FIG. 10. Response axis
53 shows amount of response as position is moved along an X-axis 51 and an
X-axis 52. Estimated path of steepest ascent response 55 (Po-est) may be
contrasted with path of actual response 56 (Po-act).
Step 1d: Alignment and sealing station 3 is placed at the measured
coordinate location with the greatest actual response (Po-act). From Table
2 it is seen that this is where X-axis equals +15 micrometers and Y-axis
equals -9.48 micrometers. The chosen coordinate point (15,-9.48,0) is now
re-labelled (0,0,0) in preparation for step 2.
Step 2a: Having found the high yield response region, measurements are made
in order to facilitate use of a second order composite approximation of
the alignment which yields the highest response. For a designate of the
three squared factorial design nine test condition runs are required.
Example values for the alignment of the X-axis and the Y-axis coordinates,
the coded form of each alignment and the measured response at each of the
nine alignments are shown in Table 3 below:
TABLE 3
______________________________________
X-axis X1 Y-axis Y1 Response (P)
______________________________________
-5 -1 -5 -1 5.500
+5 +1 -5 -1 5.530
-5 -1 +5 +1 5.500
+5 +1 +5 +1 5.555
-7.07 -1.414 0 0 5.500
+7.07 +1.414 0 0 5.555
0 0 -7.07 -1.414
5.515
0 0 +7.07 +1.414
5.530
0 0 0 0 5.555
______________________________________
These measured points are shown in FIG. 11 as points 61, 16, 63, 64, 65,
66, 67, 68 and 69.
Step 2b: Computer 1 computes the response surface of the second order
polynomial given by equation 4 below:
Po=b0=b1X1=b2Y1=b11(X1.sup.2)=b22(Y1.sup.2)=b12(X1Y1) (4)
Coefficients b0, b1, b11, b22, and b12 are calculated by the least squares
method using matrix equation (2) above. In equation (2), matrix A is as
follows:
Matrix A
______________________________________
Matrix A
______________________________________
1.0 -1.0 -1.0 1.0 1.0 1.0
1.0 1.0 -1.0 1.0 1.0 -1.0
1.0 -1.0 1.0 1.0 1.0 -1.0
1.0 1.0 1.0 1.0 1.0 1.0
1.0 -1.414 0.0 2.0 0.0 0.0
1.0 1.414 0.0 2.0 0.0 0.0
1.0 0.0 -1.414 0.0 2.0 0.0
1.0 0.0 1.414 0.0 2.0 0.0
1.0 0.0 0.0 0.0 0.0 0.0
______________________________________
Matrix A' is the transpose of Matrix A. Using the values in table 3 above,
calculated values for the coefficients are as follows: b0 equals +5.555;
b1 equals +1.020; b2 equals +0.006; b11 equals -0.015; b22 equals -0.017;
and, b12 equals +0.006.
Step 2c: Computer 1 determines from equation (4) the coordinate values at
which a maximum or minimum response occurs. This is done using equation
(5) and equation (6) which respectively are the derivative of Po with
respect to X1 and the derivative of Po with respect to Y1:
d(Po)/d/(X1)=b1=2*b11(X1)=b12(Y1) (5)
d(Po)/d/(Y1)=b2=2*b22(Y2)=b12(X1) (6)
The derivative in equation (5) is set to zero and the derivative in
equation (6) is set to zero. The simultaneous equations are then solved to
find the values X1 and Y1 for maximum or minimum Po.
Step 2d: Computer 1 substitutes the values for X1 and Y1 found in Step 2c
into equation (4) to determine the estimated response (Po-est). Computer 1
performs six confirmatory tests by measuring actual responses (Po-act) at
positions in the X-Y plane. Po-est versus Po-act are given in table 4
below:
TABLE 4
______________________________________
X-axis
X1 Y-axis Y1 Po-est
Po-act
______________________________________
0 0 0 0 5.55 5.56
+0.76 +0.152 +0.31 +0.061 5.56 5.57
+1.52 +0.303 +0.61 +0.122 5.56 5.57
+2.27 +0.455 +0.92 +0.184 5.56 5.57
+3.03 +0.606 +1.22 +0.245 5.56 5.58
+3.79 +0.758 +1.53 +0.306 5.56 5.58
______________________________________
The alignment that yields the peak response is selected. In this case it
would be coordinate (3.03, 1.22,0) which is also (X-peak,Y-peak,0).
Computer 1 resets this coordinate to be coordinate (0,0,0).
Step 3: Computer 1 varies the motor's Z-coordinate position until the
optimum Z-coordinate position (i.e., the Z-coordinate position which
results in the highest response) is found.
Step 4a: Upon finding Z-optimum, measurements for a second order
approximation are again obtained. The response at nine alignments are
obtained. Example values for the alignment of the X-axis and the Y-axis
coordinates, the coded form of each alignment and the measured response at
each of the nine alignments are shown in Table 5 below:
TABLE 5
______________________________________
X-axis X1 Y-axis Y1 Response
______________________________________
-3 -1 -3 -1 9.075
0 0 -3 -1 9.065
+3 +1 -3 -1 9.060
-3 -1 0 0 9.070
0 0 0 0 9.070
+3 +1 0 0 9.050
-3 -1 +3 +1 9.070
0 0 +3 +1 9.060
+3 +3 +3 +1 9.035
______________________________________
Step 4b: Computer 1 then computes the response surface of the second order
polynomial given by equation (4). Coefficients b0, b1, b11, b22, and b12
are calculated by the least squares method using matrix equation (2)
above. Matrix A is as follows:
Matrix A
______________________________________
Matrix A
______________________________________
1.0 -1.0 -1.0 1.0 1.0 1.0
1.0 0.0 -1.0 0.0 1.0 0.0
1.0 1.0 -1.0 1.0 1.0 -1.0
1.0 -1.0 0.0 1.0 0.0 0.0
1.0 0.0 0.0 0.0 0.0 0.0
1.0 1.0 0.0 1.0 0.0 0.0
1.0 -1.0 1.0 1.0 1.0 -1.0
1.0 0.0 1.0 0.0 1.0 0.0
1.0 1.0 1.0 1.0 1.0 1.0
______________________________________
Using the values in table 5 above, calculated values for the coefficients
are as follows: b0 equals 9.067; b1 equals -0.012; b2 equals -0.006; b11
equals -0.005; b22 equals -0.002; and b12 equals -0.005.
Step 4c: Computer 1 determines from equation (4) the coordinate values at
which a maximum or minimum response occurs. This is again done using
equation (5) and (6). The derivative in equation (5) is set to zero and
the derivative in equation (6) is set to zero. The simultaneous equations
are solved in order to find the values X1 and Y1 for maximum or minimum
Po.
Step 4d: Computer 1 substitutes the values for X1 and Y1 found in Step 4c
into equation 4 to determine the estimated response (Po-est). Computer 1
performs six confirmatory tests by measuring actual responses (Po-act) at
positions in the X-Y plane. Po-est versus Po-act are given in table 6
below:
TABLE 6
______________________________________
X-axis
X1 Y-axis Y1 Po-est
Po-act
______________________________________
0 0 +2.1E-12 0 9.07 9.06
-1.17 +0.152 -4.3E-12 0 9.07 9.06
-2.3 +0.303 -6.5E-12 0 9.07 9.07
-3.5 +0.455 -6.7E-12 0 9.07 9.06
-4.67 +0.606 -8.7E-12 0 9.07 9.07
-5.83 +0.758 -1.1E-11 0 9.07 9.07
______________________________________
The alignment that yields the peak response is selected. In this case it
would be coordinate (-2.33, 0,Z-optimum) which is also
(X-optimum,Y-optimum, Z-optimum).
The method of the present invention compares favorably with the
one-variable-at a time method. In a typical one-variable-at-a time method,
for alignment in the x-y coordinate plane 80 coordinate points are
measured in the crude alignment (before the Z direction alignment) and
another 80 coordinate points are measured in the final alignment (after
the Z direction alignment). Using the method of the present invention,
however, crude alignment in the X-Y plane was done measuring the response
at only 35 coordinate points. The final alignment in the X-Y plane was
performed measuring only 15 coordinate points.
For a two cubed factorial, the surface response method can be used with
only a few changes. There are essentially still two basic steps: (1)
Measurements are made from which the path of steepest ascent is determined
by first order design model approximation. (2) Further measurements are
made within a more limited region determined by the first order design
model approximation. From the further measurements second order modelling
is done to approximate the alignment resulting in a maximum response.
The first order equation (1) is changed to equation (7) below:
Po=b0=b1X1=b2Y1=b3Z1 (7)
The second order equation 1 is changed to equation (8)
Po=b0=b1X1=b2Y1=b3Z3=b12(X1Y1)=b13(X1Z1) =b23(Y1Z1)=b123X1Y1Z1 (8)
The coded settings of each alignment for the second order composite design
are as in table 7 below:
TABLE 7
______________________________________
X1 Y1 Z1
______________________________________
-1 -1 -1
1 -1 -1
-1 1 -1
1 1 -1
-1 -1 1
1 -1 1
-1 1 1
1 1 1
-1.682 0 0
1.682 0 0
0 -1.682 0
0 1.682 0
0 0 -1.682
0 0 1.682
0 0 0
______________________________________
Matrix A for the second order composite equation is given below:
Matrix A
______________________________________
Matric A
______________________________________
1 -1 -1 -1 1 1 1 -1
1 1 -1 -1 -1 -1 1 1
1 -1 1 -1 -1 1 -1 1
1 1 1 -1 1 -1 -1 -1
1 -1 -1 1 1 -1 -1 1
1 1 -1 1 -1 1 -1 -1
1 -1 1 1 -1 -1 1 -1
1 1 1 1 1 1 1 1
1 -1.682 0 0 0 0 0 0
1 1.682 0 0 0 0 0 0
1 0 -1.682 0 0 0 0 0
1 0 1.682 0 0 0 0 0
1 0 0 -1.682
0 0 0 0
1 0 0 1.682 0 0 0 0
1 0 0 0 0 0 0 0
______________________________________
General statistical information related to the present invention may be
found in Box, Hunter & Hunter. Statistics for Experimenters. Chapters 14
and 15.
Top